The part of systemd that provides pid 1 is relatively small and self contained. Sure there is systemd-networkd, systemd-tmpfiles, systemd-logind, etc, but those all speak to systemd pid 1 via dbus mostly. Anything that wanted to pretend to be one of the various daemons can simply speak the dbus protocol. In fact, this was one of the discussion points in the Debian systemd debate recently.
> Anything that wanted to pretend to be one of the various daemons can simply speak the dbus protocol.
The internal DBus API that these components use is not stable and will not be stable for the foreseeable future; according to the systemd developers, parts of systemd cannot be re-implemented independently [1]. That effectively makes systemd a monolithic system--its components are tightly-coupled and cannot be readily swapped out.
Note that a system can be both monolithic and modular--systemd, X.org, and the Linux kernel being examples. Systemd gets a bad rap about it because the components that make up systemd were once loosely-coupled, and the need for the new coupling is unjustified.
Who defines the dbus protocol? Where's the relevant independent standards body?
If we were talking about web browsers, most people would rightfully be livid, as we've seen browsers that decided to go off and do their own thing before, and we're still suffering from them.
When has Linux ever been beholden to an independent standards body? This is an operating built on de facto standardization, not top-down bureaucracy.
The W3 is a "tell me where the people are going so I can lead them there" type of organization. Almost every standardized feature in HTML was first implemented as an extension in some browser. When they did conceive of RDF in a committee it received only sparse adoption. We're no closer to a semantic web today than we were in 2002.
Linux has only adhered to standards where the standards are actually useful. And it has never waited for an official standard before venturing in a new direction.
> When has Linux ever been beholden to an independent standards body?
It hasn't needed to be because nobody's tried to replace so much of userspace with such a tightly integrated set of programs before, and we've not had programs and services assume that such a mess is being used before.
> Linux has only adhered to standards where the standards are actually useful.
I would argue that at the point that a number of programs start to depend on complex functionality, where the API is defined by a grand total of one project, some form of standardisation is required.
Otherwise, it is nearly impossible for others to reimplement the API correctly, meaning that we will never have meaningful competition within this space without breaking half the services anyone runs. That is not a good thing.
Before systemd, I could replace any component of my system independently, and everything would keep working. To a very significant extent, I could even replace my kernel with a BSD, recompile everything, and it all keeps on working! Now, I can no longer do that, because of a variety of packages which have hard dependencies on systemd where previously they worked fine without it.
